ESM-LSTM based optimisation model for weather prediction

This study introduces an optimized prediction model based on the integration of the Event Synchronization Method (ESM) and Long Short-Term Memory (LSTM). The primary objective is to address the limitations inherent in traditional forecasting models, particularly in understanding the intricate mechanisms of complex weather systems and handling uncertainty factors. The proposed model aims to enhance both prediction accuracy and true positive rates. The research methodology involves an initial rough-graining process applied to rainfall data, followed by the construction of a complex rainfall network using ESM. Subsequently, this network is seamlessly integrated with an optimized LSTM model, complemented by a rough K-nearest neighbors classifier for comprehensive weather predictions. The results of the comprehensive evaluation indicate a notable $20 \%$ improvement in rainfall true positive rates, while maintaining accuracy levels comparable to those of traditional models. This optimized model introduces innovative perspectives and methodologies to the field of weather prediction, providing valuable insights for researchers in related domains.